In 2025, BYD leads the charge in the "Popularization of Intelligent Driving" | Exclusive from 36Kr

The intelligent driving industry is on the brink of an explosion.

"Our mass production scale this year might increase by fourfold." An executive of a leading intelligent driving company informed 36Kr Auto.

2025 is also the first year of delivery for Horizon Journey 6 series, and the shipment quantity is anticipated to directly exceed 1 million units. "It will assist more than 100 vehicle models with medium and high-level intelligent driving functions to be launched in the market, and the shipment volume of its Journey family will surpass 10 million within the year," Yu Kai, the founder of Horizon, also stated recently.

The explosive growth of the intelligent driving companies' business directly corresponds to the demand of automakers. Compared to last year, the main driving force of intelligent driving is no longer the new forces such as Huawei, Xiaopeng, Li Auto, NIO, Xiaomi, and Zeekr, but has become the established automakers represented by BYD.

36Kr Auto exclusively learned that starting from this year, some models of BYD will gradually deliver its self-developed medium and high-level intelligent driving solutions, including high-speed pilot and urban pilot, and the self-developed intelligent driving team of approximately 1,300 people is rushing for the final delivery.

"The most crucial task for BYD at present is to achieve equal access to intelligent driving and simultaneously realize the implementation of self-developed algorithms," an informed source said.

According to 36Kr Auto, BYD's high-level intelligent driving self-developed project started in September last year, and now it is "like the Long March, only one step away from the finish." And this solution will be largely equipped in medium and low-priced models that were previously not associated with medium and high-level intelligent driving.

In fact, in August last year, Yang Dongsheng, the head of BYD's New Technology Research Institute, said in an interview with 36Kr Auto that in the future, high-level intelligent driving functions will be further decentralized, "even cars under 100,000 yuan should possess them."

Many industry insiders told 36Kr Auto that they did not anticipate BYD to move so rapidly. If the self-developed solution can be delivered in the first half of the year, it is basically a blitz, and many peers are not prepared yet. Many suppliers told 36Kr Auto that the delivery expectations for the high-speed NOA solution requirements for 100,000 - 150,000 yuan vehicle models they received are in 2026.

The aftershocks of the price war in 2024 obviously continue into 2025, and its external manifestation has changed from blatant price competition to configuration competition. Sources stated, "Even for BYD's 100,000-yuan-level models, they advocate equipping 7/11 cameras." Not long ago, in the 390th batch of new vehicle catalogs announced by the Ministry of Industry and Information Technology, nearly 20 models of BYD's Dynasty Network and Ocean Network are equipped with hardware such as three-eye vision.

According to industry predictions, BYD's full-year sales in 2025 are expected to be 5.5 million. And 36Kr Auto exclusively learned that BYD has placed million-level chip orders with both NVIDIA and Horizon.

If the intelligent driving deliveries of automakers such as BYD proceed smoothly, the automotive industry will usher in the delivery of millions of medium and high-level intelligent driving vehicles in 2025. This is almost the total delivery of intelligent driving companies that focus on intelligent driving, such as Li Auto, NIO, Xiaopeng, and Huawei.

Insiders disclosed that BYD is also making intensive preparations for this, "The corresponding data closed-loop work and after-sales plans have already been in place."

In the past year, the intelligent driving arena was basically supported by new forces such as Huawei, Li Auto, NIO, and Xiaopeng, with few established automakers gaining a foothold. However, the efforts of established automakers such as BYD in the mass production of intelligent driving are likely to promote the rapid popularization of medium and high-level intelligent driving. This is another major industrial change in the automotive industry after electrification.

The Main Force for the Popularization of Intelligent Driving: From New Automakers to Established Automakers

Last year, BYD's intelligent driving team underwent multiple integrations and gradually intensified the self-developed intelligent driving.

Based on the publicly available information of BYD, BYD's intelligent driving is currently mainly divided into three different solutions: DiPilot 100, DiPilot 300, and DiPilot 600, with computing powers of less than 100 Tops, 250 Tops, and 500 Tops, respectively.

DiPilot 100 mainly supports functions such as high-speed NOA, which is also the key direction of BYD's self-developed delivery at present.

Both DiPilot 300 and DiPilot 600 support urban NOA, with computing power solutions of 1 Orin X and 2 Orin X, respectively; Another difference is the number of lidars. For example, the two different platforms corresponding to Tengshi N7 and Yangwang U8 are equipped with 2 and 3 lidars, respectively.

In addition to BYD, established automakers such as Great Wall, Geely, Changan, and Chery have all entered different stages of self-developed intelligent driving.

For instance, Great Wall Motors previously collaborated with the intelligent driving company DeepRoute.ai, which provided a high-level intelligent driving solution for Great Wall's Wey Coffee 01 model. Nearly half of the company's more than 700 employees are working on the Great Wall project.

However, Great Wall's self-developed intelligent driving team is also vigorously recruiting. An informed person told 36Kr Auto that Great Wall's intelligent driving is still expanding the team and will establish an office in California, USA. At the same time, in order to accelerate the intelligent driving penetration rate of its vehicle models, Great Wall is also in contact with intelligent driving companies such as Zhuoyu (i.e., DJI Auto) and QCraft for medium-level intelligent driving mass production projects.

After a bold adjustment, Geely has formed two intelligent driving routes internally: Zeekr focuses on high-level, and Geely Automobile focuses on high-value (low-level) intelligent driving solutions.

Changan Automobile's intelligent driving has also established a 1,000-person intelligent driving team, and the self-developed solution plans to transition from low-level to high-level intelligent driving; Chery Automobile also expects that more than ten models of its four major brands will adopt the self-developed flagship high-level intelligent driving NOA product for mass production and listing in 2025, and the configuration rate of the main sales models will exceed 50%.

It is not difficult to see that in 2025, established automakers have bet more chips on intelligent driving. However, the progress of self-developed intelligent driving of most established automakers is still at the high-speed NOA stage or an earlier stage, and they cannot directly participate in the popular intelligent driving technology competitions such as urban NOA and end-to-end.

But from the perspective of penetration rate, the popularizers of intelligent driving may no longer be the new forces, but the established automakers with a larger sales volume. The game between self-research and procurement of established automakers represented by BYD will also become the biggest variable in the intelligent driving industry.

The Resurgence of Automakers' Self-Developed Intelligent Driving: An Opportunity or a Trap?

Since its first launch in China in 2019 until now, the consumer group of Model Y has changed. Many

Why did the self-developed intelligent driving of established automakers suddenly explode in 2025?

This is due to the increasingly evident commercial value of intelligent driving. Last year, under the exploration of three business models such as Smart Selection Vehicle, Huawei's Vehicle Business Unit achieved nearly 10 billion in revenue in half a year; Li Auto also made a breakthrough with intelligent driving and achieved an intelligent driving vehicle penetration rate of more than 50%.

For established automakers who wish to join the wave of intelligence, there are two paths before them: self-research and procurement.

The effect of adopting the intelligent driving supplier solution is immediate. For example, after BYD's Tengshi N7 is equipped with Momenta's end-to-end solution, BYD's Fangchengbao Leopard 8 is equipped with Huawei ADS 3.0 intelligent driving, and Great Wall's Wey Coffee 01 is equipped with DeepRoute.ai's solution, they can quickly catch up with the industry's first-tier level.

However, last year, the price of vehicle models with urban NOA function in the industry was basically above 200,000 yuan. A management of a leading intelligent driving company predicted to 36Kr Auto that in 2025, the urban NOA function will become a standard feature for new cars priced at 150,000 yuan, and by 2026, the urban NOA function will drop to the 100,000-yuan level.

As for high-speed NOA, it may be commonly equipped in 100,000-yuan-level vehicle models this year. With the further popularization of intelligent driving, established automakers have to balance between scale cost and mass production speed.

An intelligent driving supplier told 36Kr Auto that the time for its intelligent driving function to adapt to new vehicle models has been shortened to 3 - 6 months. If automakers adopt the supplier solution, it is not difficult to achieve the rapid popularization of high-speed NOA.

But the fundamental reason for established automakers to decide to self-develop intelligent driving lies in scale and cost.

According to 36Kr Auto, if automakers adopt the supplier solution, they generally need to pay tens of millions of platform development fees and vehicle adaptation fees, and at the same time, they also need to pay the license authorization fee for the intelligent driving software for each vehicle sold. Taking the high-speed NOA function as an example, the license fee for each vehicle that automakers need to pay is about 1,000 yuan.

In other words, the better the sales of automakers' vehicle models, the higher the overall license fee. For example, the software authorization fee for 10,000 vehicles with high-speed NOA is tens of millions of yuan, and the software authorization fee for 100,000 vehicles will exceed 100 million yuan.

For automakers with annual sales of millions, this is an account that needs to be considered. In contrast, self-developing high-speed NOA may be a more cost-effective path.

An intelligent driving person told 36Kr Auto that after several years of industry exploration, functions such as high-speed NOA are actually relatively standardized and mature, and the product experience of the first and second tiers in the industry is not so disparate; The difficulty of self-research by automakers is not too great, and it is relatively easy to achieve results. For example, BYD's self-developed high-speed NOA project, which started in September last year, is now close to the end of development.

Moreover, the boundaries between the current automotive industry and fields such as AI and embodied intelligence are becoming increasingly blurred, and some automakers regard artificial intelligence companies as their development vision. When intelligent driving technology is increasingly bound to AI, big data, and computing power, automakers find it difficult to resist the temptation of self-research and mastering core technologies.

But the bottleneck of automakers' self-developed intelligent driving may also emerge.

At present, the high-speed NOA and lightweight urban NOA functions self-developed by established automakers are mostly based on the traditional rule-based intelligent driving solution.

Routes like "end-to-end", where intelligent driving is deeply integrated with AI and large models, are a greater leap in intelligent driving technology and a newer species, and established automakers are slightly struggling. According to 36Kr Auto, BYD is also exploring the end-to-end intelligent driving solution, but the landing time will be at least until 2026.

If the overall "end-to-end" solution in the industry converges in the next two years, then the manpower and resources invested by established automakers are expected to catch up; but if the end-to-end solution accelerates the integration with AI, robots, etc., and creates a larger gap, then the self-developed intelligent driving of established automakers undoubtedly requires a stronger acceleration.

A management of an intelligent driving company told 36Kr Auto that automakers can definitely make something if they self-develop, but the more difficult and the more investment is required as the intelligent driving progresses. Now, an annual investment of several billion yuan can reach the quasi-first-tier level, but as it progresses, it may require tens of billions or even higher investment to achieve the same level. "The gap may be widening, not narrowing."

He believes that if the self-developed product is not at the first-tier level, it will not be beneficial for selling cars, and then some automakers may abandon the path of self-developed intelligent driving.